Create

Model Context Protocol

Kverton for AI assistants

Claude, ChatGPT, Cursor and other assistants can work with your Kverton directly through an MCP server: create a dynamic code, change where it leads, render it, read statistics. You say what you want; the assistant calls the same API as your own systems.

What it is

Model Context Protocol is an open standard through which AI assistants gain tools. The Kverton MCP server is a small program (npm package) that runs on your machine and translates the assistant's requests into Kverton API calls under your key. Nothing is installed on Kverton's server; data flows only between your assistant and your team.

“Create a dynamic code for the autumn menu at example.com/menu-autumn, render it as SVG, and tell me tomorrow how many people scanned it in Prague.”

The assistant calls kverton_create_code, kverton_render_svg and, the next day, kverton_code_stats — and answers with a number, not a link to a manual.

Two-minute setup

  1. Create an API key in Kverton: Teams → team → API keys. A read-only key is enough for statistics; use read + write to create and change codes.
  2. Add the server to your assistant (Node.js 18+ required on your machine):

Claude Desktop — claude_desktop_config.json

{
  "mcpServers": {
    "kverton": {
      "command": "npx",
      "args": ["-y", "@kverton/mcp"],
      "env": { "KVERTON_API_KEY": "kv_XXXXXXXX_…" }
    }
  }
}

Claude Code

claude mcp add kverton \
  -e KVERTON_API_KEY=kv_XXXXXXXX_… \
  -- npx -y @kverton/mcp

Cursor, Windsurf, ChatGPT and others

Every MCP client has an “add server” setting — fill in the command and the environment variable: npx -y @kverton/mcp, KVERTON_API_KEY.

The server runs locally over stdio. Requires a plan with API access (Pro or Studio); calls count toward the team's daily quota.

What the assistant can do

Thirteen tools, each exactly one API endpoint. Nothing extra, nothing hidden.

ToolAPIScope
kverton_me GET /me read
kverton_list_codes GET /codes read
kverton_create_code POST /codes write
kverton_get_code GET /codes/{id} read
kverton_update_code PATCH /codes/{id} write
kverton_delete_code DELETE /codes/{id} write
kverton_code_stats GET /codes/{id}/stats read
kverton_code_history GET /codes/{id}/history read
kverton_rollback_code POST /codes/{id}/rollback write
kverton_list_groups GET /groups read
kverton_render_svg POST /render read
kverton_list_webhooks GET /webhooks read
kverton_create_webhook POST /webhooks write

Safety and boundaries

  • The assistant has exactly the rights of the key — a read-only key cannot change or delete anything.
  • The key stays in your assistant's configuration; Kverton never shows it again and you can revoke it anytime in team settings.
  • Every call appears in the key's usage overview (how many, when, what for) and counts toward the plan quota.
  • Deleting a code is irreversible — the assistant is instructed to prefer pausing; still, consider a read-only key for everyday questions.